A numerical approximation method for RSS-based user positioning in distributed massive MIMO
K. N. R. Surya Vara Prasad, Ekram Hossain, Vijay K. Bhargava · 2017
We consider user positioning in distributed massive multiple-input multiple-output (MIMO) systems based on uplink received signal strength (RSS) recorded at the base station. We take a machine learning approach, wherein, we train a Gaussian process regression (GP) model with uplink RSS vectors for several known training user locations and use the trained GP model for predicting locations of test users from their uplink RSS vectors. When the training RSS is noise-free and the test RSS is noisy due to shadowing from the wireless channel, we observe that the conventional GP method provides unrealistically small 2σ error-bars on the predicted locations. To address this shortcoming, we propose NAGP — a numerical approximation based GP method which employs averaging over Monte-Carlo samples to improve the 2σ error-bar estimates. Simulation studies confirm the superior performance offered by the NAGP method.